[gmx-users] Continuation 1) after crash and 2) in general

David van der Spoel spoel at xray.bmc.uu.se
Tue Dec 11 20:06:22 CET 2001

On Tue, 11 Dec 2001, Marc Baaden wrote:

>in a recent message spoel at xray.bmc.uu.se said:
>>> Kill the simulation and restart it from where the error in the xtc
>>> started (tpbconv -s topol -f traj -o cont -time 10000000) Fill in the
>>> right number for the time there. 
>I tried the tbpconv on a crash that happened recently. My only concern
>is, does it properly take into account shuffling ? Eg if the initial
>simulation was shuffled, will the continuation run be shuffled in the
>same way ?
I'm *quite* sure. The trr is shuffled and the tpr as well, since shuffling
occurs in grompp. Try it with a small membrane simulation, it will
certainly break if it's wrong...

>And a more general question: if I want to split my simulation of let's
>say 5 ns into chunks of 1 ns, what do I need in order to obtain the
>same thing as in one long simulation, eg no "discontinuities" at the
>simulation ends (1 ns intervals).
Basically not more than tpbconv, but there is one but. If you use
T-coupling the trajectories are not identical to what a single one would
be. There is nothing we can do about it right now, but it is on the

>I mean, is it sufficient to write out coords+velocities at the end ?
>(Because due to the integration algorithm they should be t/2 shifted
>during a run, but what happens when you write them out ?)
V is written at half time steps.

>Is there a kind of model "restart-mdp" file ?
tpbconv -s topol.tpr -f traj.trr -o conv -extend 1000

>And is it always necessary to build a complete topology when you
>restart ? With eg AMBER you can use the same topology and just indicate
>the new coords/vels to start from and which time it is.
Right now tpbconv is as easy as it gets...

Groeten, David.
Dr. David van der Spoel, 	Biomedical center, Dept. of Biochemistry
Husargatan 3, Box 576,  	75123 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://zorn.bmc.uu.se/~spoel

More information about the gromacs.org_gmx-users mailing list